1 Boolean Logic Such simple things, And we make of them something so complex it defeats us, Almost. —John Ashbery (b. 1927), American poet Every digital device—be it a personal computer, a cellular telephone, or a network router—is based on a set of chips designed to store and process information. Al- though these chips come in different shapes and forms, they are all made from the same building blocks: Elementary logic gates. The gates can be physically imple- mented in many different materials and fabrication technologies, but their logical behavior is consistent across all computers. In this chapter we start out with one primitive logic gate—Nand—and build all the other logic gates from it. The result is a rather standard set of gates, which will be later used to construct our computer’s processing and storage chips. This will be done in chapters 2 and 3, respectively. All the hardware chapters in the book, beginning with this one, have the same structure. Each chapter focuses on a well-defined task, designed to construct or inte- grate a certain family of chips. The prerequisite knowledge needed to approach this task is provided in a brief Background section. The next section provides a complete Specification of the chips’ abstractions, namely, the various services that they should deliver, one way or another. Having presented the what, a subsequent Implemen- tation section proposes guidelines and hints about how the chips can be actually implemented. A Perspective section rounds up the chapter with concluding com- ments about important topics that were left out from the discussion. Each chapter ends with a technical Project section. This section gives step-by-step instructions for actually building the chips on a personal computer, using the hardware simulator supplied with the book. This being the first hardware chapter in the book, the Background section is somewhat lengthy, featuring a special section on hardware description and simulation tools. 8 Chapter 1 1.1 Background This chapter focuses on the construction of a family of simple chips called Boolean gates. Since Boolean gates are physical implementations of Boolean functions,we start with a brief treatment of Boolean algebra. We then show how Boolean gates implementing simple Boolean functions can be interconnected to deliver the func- tionality of more complex chips. We conclude the background section with a descrip- tion of how hardware design is actually done in practice, using software simulation tools. 1.1.1 Boolean Algebra Boolean algebra deals with Boolean (also called binary) values that are typically labeled true/false, 1/0, yes/no, on/off, and so forth. We will use 1 and 0. A Boolean function is a function that operates on binary inputs and returns binary outputs. Since computer hardware is based on the representation and manipulation of binary values, Boolean functions play a central role in the specification, construction, and optimization of hardware architectures. Hence, the ability to formulate and analyze Boolean functions is the first step toward constructing computer architectures. Truth Table Representation The simplest way to specify a Boolean function is to enumerate all the possible values of the function’s input variables, along with the function’s output for each set of inputs. This is called the truth table representation of the function, illustrated in figure 1.1. The first three columns of figure 1.1 enumerate all the possible binary values of the n function’s variables. For each one of the 2 possible tuples v1 ...vn (here n ¼ 3), the last column gives the value of f ðv1 ...vnÞ. Boolean Expressions In addition to the truth table specification, a Boolean function can also be specified using Boolean operations over its input variables. The basic Boolean operators that are typically used are ‘‘And’’ (x And y is 1 exactly when both x and y are 1) ‘‘Or’’ (x Or y is 1 exactly when either x or y or both are 1), and ‘‘Not’’ (Not x is 1 exactly when x is 0). We will use a common arithmetic-like notation for these operations: x Á y (or xy) means x And y, x þ y means x Or y, and x means Not x. To illustrate, the function defined in figure 1.1 is equivalently given by the Boolean expression f ðx; y; zÞ¼ðx þ yÞz. For example, let us evaluate this expression on the 9 Boolean Logic xyz fðx; y; zÞ 000 0 001 0 010 1 011 0 100 1 101 0 110 1 111 0 Figure 1.1 Truth table representation of a Boolean function (example). inputs x ¼ 0, y ¼ 1, z ¼ 0 (third row in the table). Since y is 1, it follows that x þ y ¼ 1 and thus 1 Á 0 ¼ 1 Á 1 ¼ 1. The complete verification of the equivalence between the expression and the truth table is achieved by evaluating the expression on each of the eight possible input combinations, verifying that it yields the same value listed in the table’s right column. Canonical Representation As it turns out, every Boolean function can be expressed using at least one Boolean expression called the canonical representation. Starting with the function’s truth table, we focus on all the rows in which the function has value 1. For each such row, we construct a term created by And-ing together literals (variables or their negations) that fix the values of all the row’s inputs. For example, let us focus on the third row in figure 1.1, where the function’s value is 1. Since the variable values in this row are x ¼ 0, y ¼ 1, z ¼ 0, we construct the term xyz. Fol- lowing the same procedure, we construct the terms xyz and xyz for rows 5 and 7. Now, if we Or-together all these terms (for all the rows where the function has value 1), we get a Boolean expression that is equivalent to the given truth table. Thus the canonical representation of the Boolean function shown in figure 1.1 is f ðx; y; zÞ¼ xyz þ xyz þ xyz. This construction leads to an important conclusion: Every Boolean function, no matter how complex, can be expressed using three Boolean operators only: And, Or, and Not. Two-Input Boolean Functions An inspection of figure 1.1 reveals that the number of Boolean functions that can be defined over n binary variables is 22 n . For example, the sixteen Boolean functions spanned by two variables are listed in figure 1.2. These functions were constructed systematically, by enumerating all the possible 4-wise com- binations of binary values in the four right columns. Each function has a conventional 10 Chapter 1 x 0011 Function y 0101 Constant 0 0 0000 And x Á y 0001 x And Not yxÁ y 0010 xx0011 Not x And y x Á y 0100 yy0101 Xor x Á y þ x Á y 0110 Or x þ y 0111 Nor x þ y 1000 Equivalence x Á y þ x Á y 1001 Not y y 1010 If y then xxþ y 1011 Not x x 1100 If x then y x þ y 1101 Nand x Á y 1110 Constant 1 1 1111 Figure 1.2 All the Boolean functions of two variables. name that seeks to describe its underlying operation. Here are some examples: The name of the Nor function is shorthand for Not-Or: Take the Or of x and y, then negate the result. The Xor function—shorthand for ‘‘exclusive or’’—returns 1 when its two variables have opposing truth-values and 0 otherwise. Conversely, the Equivalence function returns 1 when the two variables have identical truth-values. The If-x-then-y function (also known as x ! y,or‘‘x Implies y’’) returns 1 when x is 0 or when both x and y are 1. The other functions are self-explanatory. The Nand function (as well as the Nor function) has an interesting theoretical property: Each one of the operations And, Or, and Not can be constructed from it, and it alone (e.g., x Or y ¼ðx Nand xÞ Nand ðy Nand yÞ. And since every Boolean function can be constructed from And, Or, and Not operations using the canonical representation method, it follows that every Boolean function can be constructed from Nand operations alone. This result has far-reaching practical implications: Once we have in our disposal a physical device that implements Nand, we can use many copies of this device (wired in a certain way) to implement in hardware any Boolean function. 11 Boolean Logic 1.1.2 Gate Logic A gate is a physical device that implements a Boolean function. If a Boolean function f operates on n variables and returns m binary results (in all our examples so far, m was 1), the gate that implements f will have n input pins and m output pins. When we put some values v1 ...vn in the gate’s input pins, the gate’s ‘‘logic’’—its internal structure—should compute and output f ðv1 ...vnÞ. And just like complex Boolean functions can be expressed in terms of simpler functions, complex gates are com- posed from more elementary gates. The simplest gates of all are made from tiny switching devices, called transistors, wired in a certain topology designed to effect the overall gate functionality. Although most digital computers today use electricity to represent and transmit binary data from one gate to another, any alternative technology permitting switch- ing and conducting capabilities can be employed. Indeed, during the last fifty years, researchers have built many hardware implementations of Boolean functions, includ- ing magnetic, optical, biological, hydraulic, and pneumatic mechanisms.